This is an automated email from the ASF dual-hosted git repository.
alamb pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/datafusion-site.git
The following commit(s) were added to refs/heads/main by this push:
new 2e26e44 Blog post on Parquet pruning in datafusion (#60)
2e26e44 is described below
commit 2e26e44e2d91a5c39154f8ad3fbb7379c8b91a50
Author: Xiangpeng Hao <[email protected]>
AuthorDate: Thu Mar 20 11:38:28 2025 -0500
Blog post on Parquet pruning in datafusion (#60)
* new blog post
* Apply suggestions from code review
Co-authored-by: Andrew Lamb <[email protected]>
* improve writing
* update
* Update date
* Capitalization consistency
---------
Co-authored-by: Andrew Lamb <[email protected]>
---
content/blog/2025-03-20-parquet-pruning.md | 118 +++++++++++++++++++++
content/images/parquet-pruning/filter-pushdown.jpg | Bin 0 -> 376764 bytes
content/images/parquet-pruning/read-parquet.jpg | Bin 0 -> 284097 bytes
3 files changed, 118 insertions(+)
diff --git a/content/blog/2025-03-20-parquet-pruning.md
b/content/blog/2025-03-20-parquet-pruning.md
new file mode 100644
index 0000000..ffbdfed
--- /dev/null
+++ b/content/blog/2025-03-20-parquet-pruning.md
@@ -0,0 +1,118 @@
+---
+layout: post
+title: Parquet Pruning in DataFusion: Read Only What Matters
+date: 2025-03-20
+author: Xiangpeng Hao
+categories: [performance]
+---
+
+<!--
+{% comment %}
+Licensed to the Apache Software Foundation (ASF) under one or more
+contributor license agreements. See the NOTICE file distributed with
+this work for additional information regarding copyright ownership.
+The ASF licenses this file to you under the Apache License, Version 2.0
+(the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+{% endcomment %}
+-->
+
+
+_Editor's Note: This blog was first published on [Xiangpeng Hao's blog].
Thanks to [InfluxData] for sponsoring this work as part of his PhD funding._
+
+[Xiangpeng Hao's blog]: https://blog.xiangpeng.systems/posts/parquet-to-arrow/
+[InfluxData]: https://www.influxdata.com/
+<hr/>
+
+[Apache Parquet] has become the industry standard for storing columnar data,
and reading Parquet efficiently -- especially from remote storage -- is crucial
for query performance.
+
+[Apache Parquet]: https://parquet.apache.org/
+
+[Apache DataFusion] implements advanced Parquet pruning techniques to
effectively read only the data that matters for a given query.
+
+[Apache DataFusion]: https://datafusion.apache.org/
+
+Achieving high performance adds complexity.
+This post provides an overview of the techniques used in DataFusion to
selectively read Parquet files.
+
+### The pipeline
+The diagram below illustrates the [Parquet reading pipeline] in DataFusion,
highlighting how data flows through various pruning stages before being
converted to Arrow format:
+
+[Parquet reading pipeline]:
https://docs.rs/datafusion/46.0.0/datafusion/datasource/physical_plan/parquet/source/struct.ParquetSource.html```
+
+<img src="/blog/images/parquet-pruning/read-parquet.jpg" alt="Parquet pruning
pipeline in DataFusion" width="100%" class="img-responsive">
+
+
+#### Background: Parquet file structure
+As shown in the figure above, each Parquet file has multiple row groups. Each
row group contains a set of columns, and each column contains a set of pages.
+
+Pages are the smallest units of data in Parquet files and typically contain
compressed and encoded values for a specific column. This hierarchical
structure enables efficient columnar access and forms the foundation for the
pruning techniques we'll discuss.
+
+Check out [Querying Parquet with Millisecond
Latency](https://www.influxdata.com/blog/querying-parquet-millisecond-latency/)
for more details on the Parquet file structure.
+
+#### 1. Read metadata
+DataFusion first reads the [Parquet
metadata](https://parquet.apache.org/docs/file-format/metadata/) to understand
the data in the file.
+Metadata often includes data schema, the exact location of each row group and
column chunk, and their corresponding statistics (e.g., min/max values).
+It also optionally includes [page-level
stats](https://parquet.apache.org/docs/file-format/pageindex/) and [Bloom
filters](https://www.influxdata.com/blog/using-parquets-bloom-filters/).
+This information is used to prune the file before reading the actual data.
+
+[Fetching
metadata](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/datasource/physical_plan/parquet/reader.rs#L118)
requires up to two network requests: one to read the footer size from the end
of the file, and another to read the footer itself.
+
+[Decoding
metadata](https://www.influxdata.com/blog/how-good-parquet-wide-tables/) is
generally fast since it only requires parsing a small amount of data. However,
for tables with hundreds or thousands of columns, the metadata can become quite
large and decoding it can become a bottleneck. This is particularly noticeable
when scanning many small files.
+
+Reading metadata is latency-critical, so DataFusion allows users to cache
metadata through the
[ParquetFileReaderFactory](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/datasource/physical_plan/parquet/reader.rs#L39)
trait.
+
+
+#### 2. Prune by projection
+The simplest yet perhaps most effective pruning is to read only the columns
that are needed.
+This is because queries usually don't select all columns, e.g., `SELECT a FROM
table` only reads column `a`.
+As a **columnar** format, Parquet allows DataFusion to [only
read](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/datasource/physical_plan/parquet/mod.rs#L778)
the **columns** that are needed.
+
+This projection pruning happens at the column level and can dramatically
reduce I/O when working with wide tables where queries typically access only a
small subset of columns.
+
+
+#### 3. Prune by row group stats and Bloom filters
+Each row group has [basic
stats](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/physical_optimizer/pruning.rs#L81)
like min/max values for each column.
+DataFusion applies the query predicates to these stats to prune row groups,
e.g., `SELECT * FROM table WHERE a > 10` will only read row groups where `a`
has a max value greater than 10.
+
+Sometimes min/max stats are too simple to prune effectively, so Parquet also
supports [Bloom
filters](https://www.influxdata.com/blog/using-parquets-bloom-filters/).
DataFusion [uses Bloom filters when
available](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/datasource/physical_plan/parquet/opener.rs#L202).
+
+Bloom filters are particularly effective for equality predicates (`WHERE a =
10`) and can significantly reduce the number of row groups that need to be read
for point queries or queries with highly selective predicates.
+
+
+#### 4. Prune by page stats
+Parquet optionally supports [page-level
stats](https://github.com/apache/parquet-format/blob/master/PageIndex.md) --
similar to row group stats but more fine-grained.
+DataFusion implements [page
pruning](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/datasource/physical_plan/parquet/opener.rs#L219)
when the stats are present.
+
+Page-level pruning provides an additional layer of filtering after row group
pruning. It allows DataFusion to skip individual pages within a row group,
further reducing the amount of data that needs to be read and decoded.
+
+
+#### 5. Read from storage
+Now we (hopefully) have pruned the Parquet file into small ranges of bytes,
i.e., the [Access
Plan](https://github.com/apache/datafusion/blob/76a7789ace33ced54c973fa0d5fc9d1866e1bf19/datafusion/datasource-parquet/src/access_plan.rs#L86).
+The last step is to [make
requests](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/datasource/physical_plan/parquet/reader.rs#L103)
to fetch those bytes and decode them into Arrow RecordBatch.
+
+
+### Preview of coming attractions: filter pushdown
+So far we have discussed techniques that prune the Parquet file using only the
metadata, i.e., before reading the actual data.
+
+Filter pushdown, also known as predicate pushdown or late materialization, is
a technique that prunes data during scanning, with filters being generated and
applied in the Parquet reader.
+
+<img src="/blog/images/parquet-pruning/filter-pushdown.jpg" alt="Filter
pushdown in DataFusion" width="100%" class="img-responsive">
+
+Unlike metadata-based pruning which works at the row group or page level,
filter pushdown operates at the row level, allowing DataFusion to filter out
individual rows that don't match the query predicates during the decoding
process.
+
+DataFusion [implements filter
pushdown](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/core/src/datasource/physical_plan/parquet/row_filter.rs#L154)
but has [not enabled it by
default](https://github.com/apache/datafusion/blob/31701b8dc9c6486856c06a29a32107d9f4549cec/datafusion/common/src/config.rs#L382)
due to [some performance
regressions](https://github.com/apache/datafusion/issues/3463).
+
+We are working to remove the [remaining performance
issues](https://github.com/apache/arrow-rs/issues/5523#issuecomment-2429470872)
and enable it by default, which we will discuss in the next blog post.
+
+### Conclusion
+DataFusion employs a multi-step approach to Parquet pruning, from column
projection to row group stats, page stats, and potentially row-level filtering.
+Each step may reduce the amount of data to be read and processed,
significantly improving query performance.
diff --git a/content/images/parquet-pruning/filter-pushdown.jpg
b/content/images/parquet-pruning/filter-pushdown.jpg
new file mode 100644
index 0000000..23e16cf
Binary files /dev/null and b/content/images/parquet-pruning/filter-pushdown.jpg
differ
diff --git a/content/images/parquet-pruning/read-parquet.jpg
b/content/images/parquet-pruning/read-parquet.jpg
new file mode 100644
index 0000000..22b54c4
Binary files /dev/null and b/content/images/parquet-pruning/read-parquet.jpg
differ
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
